Black header with the title "How teams actually use AI in product copy workflows" above an olive-green grid of purple, blue, pink and orange blocks and circles.
Editorial

How large teams actually use AI in product copy workflows (2026 survey)

Teams are making real but uneven progress with AI workflows. Most are using chat tools or homegrown agents, and none share one source of truth for product copy yet.

AI in product copy has moved past experiments, but not as far as the conference talks and LinkedIn highlight reels suggest. We asked 130 Ditto customers how they use AI in their design and product copy work, and 32 people at 25 companies answered. 18 of 25 are large companies (over 1,000 employees).

Most teams are using AI as a chat assistant or building their own skills and agents. None have a shared system for product copy that every person and AI tool works from. And where a team sits depends less on interest than on how tightly its industry is regulated. At large banks, insurers, and medical device companies, 22% of respondents use AI agents, compared with 74% everywhere else.

Key takeaways: AI in product copy workflows

  • Most teams use AI, but haven’t systemized it yet. 48% of companies use AI as a chat assistant, and most of the rest are building their own skills, gems, and agents. None have a single shared system for product copy yet.
  • Company constraints, not interest, decide AI use. Of the respondents not using agents, more than half point to company policy or approval processes. Almost all of that group works at large companies with 1,000+ employees, and most are in regulated industries.
  • Teams sit at different stages, even within one company. Content designers often use AI as an assistant, while designers and engineers build with it.
  • Homegrown fixes have limits. Teams encode their guidelines into whatever tool they’re allowed to use, and keeping that context in sync across tools is still an open problem.
  • Work is moving closer to code. Designers are prototyping and writing copy in code tools, and industry research shows that shift is accelerating.

Who we surveyed

We surveyed 32 people at 25 companies, all Ditto customers. Most are large companies: 18 of the 25 have more than 1,000 employees, including 7 with more than 10,000. Five have 201 to 1,000 employees, and two have 200 or fewer.

About half work in regulated industries, including banking and fintech, insurance, mortgages, and medical devices. The rest span media and entertainment, enterprise software, e-learning, consumer goods, retail, telecom, gaming, food delivery, and industrial manufacturing.

19 of the 32 respondents are content designers. The rest include technical team members, leaders who sponsor Ditto within their companies, and people in marketing and operations roles.

What is the AI maturity curve for AI in content design?

We use the AI maturity curve for content design to describe how teams progress. It has four phases:

  1. No AI usage: teams aren’t using AI in their content work.
  2. Chat-based generation: teams use chat assistants to draft, edit, and brainstorm.
  3. Markdown files, skills, and gems: teams encode their content guidelines and context into tools so AI follows their rules.
  4. Infrastructure for product text: a shared system that every person and agent works from.

New to the terms? Start with our AI and agents terminology guide.

Where are teams on the curve today?

Grouping each company’s answers by phase, here’s where the respondents sit:

  1. No AI usage: 8% — Blocked by company policy
  2. Chat-based generation: 48% — Sounding board, editing, prototyping
  3. Skills, gems, and agents: 44% — Guidelines encoded into custom tools
  4. Shared infrastructure: 0% — No teams here yet

Phase 1: Blocked by policy and approval processes

The 8% of companies in phase 1 aren’t lagging out of reluctance — they plan to use AI. In one case, a global insurance company directs staff to an in-house chatbot and doesn’t allow other AI tools. At an enterprise medical device company, two respondents describe a slow organizational process and the need to work through privacy and security review first.

Phase 2: Using AI as an assistant

The largest group (48%) uses AI as an assistant for drafting, editing, ideation, and increasingly, prototyping.

“From a content point of view, I personally use AI as a sounding board for ideas… I mostly review the content that’s been created by someone else or by AI so I use it as an assistant for that.”

— Content Designer, spend management company

That matches trends in the wider industry. 96% of UX practitioners in Nielsen Norman Group’s 2026 survey of AI in UX use general AI assistants like ChatGPT, Claude, or Copilot. In UX Content Collective’s AI in Content Design 2026 report, writing (drafting, ideation, and editing) is the most common way content designers use AI day-to-day.

Phase 3: Building their own context

44% of companies have moved beyond chat to encode their rules in tooling. Here’s what that looks like:

  • A large food delivery platform built Gemini Gems trained on its UX writing guidelines, product context, and user research insights.
  • A major US media company built a style guide agent in Copilot Studio.
  • A design team at a European spend management company built a Claude agent that ingests business requirements and documentation.
  • A data governance software company uses Claude daily to draft and audit UI copy against its Ditto style guide.

Not every team calls these agents, though. A content designer at a mobile gaming company describes their customized GPTs as “assistive tools rather than autonomous agents.”

This is where the content design field at large is today. UXCC’s report found that 73% of content designers work at organizations building content-related AI tools, and 90% of those tools are built to enforce style guides and standards.

Phase 4: Shared infrastructure (teams aren’t there yet)

For teams at the head of the curve, product copy lives in one shared system that every person and AI tool works from. Instead of each skill, gem, or agent holding its own copy of the rules, they all draw from the same source: what copy already exists, what’s live, what’s been approved. When the source changes, every tool sees the change.

No organization in our survey has reached this phase, but a few are building toward it:

  • A major media company has fed its context layer into Ditto’s agent, moving its guidelines out of a single homegrown tool and into a shared system.
  • A data governance software company searches its existing copy in Ditto before writing anything new, so its AI work starts from what’s already there.
  • A global industrial equipment manufacturer uses agents to make sure designers link every string to Ditto, rather than copying text from page to page.

Teams are working toward this end state. UXCC found that 45% of content designers now spend most of their time on systems work. Figma recommends getting design systems AI-ready, so every tool works from the same foundation.

AI maturity varies by role, even within one company

Companies are rarely in a single phase. At one company, the content designer uses an AI assistant mainly to review content, while designers create initial concepts and prototypes with an agent. At another, a content designer uses Figma Make occasionally, while a technical colleague works with coding agents like Kiro and Codex.

Content designers’ AI use today tends to be less technical than that of their design and engineering counterparts. In the Nielsen Norman Group’s survey, 71% of both designers and engineers use agentic tools, but only 35% of content designers do.

Regulation and size track with where teams are on the curve

Of the respondents not clearly using agents, over half said it’s because of company policy or approval processes, such as procurement and security review. Only 15% of non-AI users cite lack of time.

Most of those blocked by policy are at large insurers, banks, and medical device companies. Meanwhile, 86% of the companies with 1,000 or fewer employees are active users.

“Making a tool compliant can take years. We stick to the basics because we have to, but that doesn’t mean there’s not a desire to adopt [AI].”

— Content Designer, global consumer goods brand

Other research shows the same pattern. Nielsen Norman Group found that agentic AI use ranges from 85% at startups to 69% at enterprises and 36% at nonprofits. Figma’s 2026 AI report notes that for some companies slow to adopt AI, industry regulations and organizational structure shape whether AI gets used at all.

What gets in the way at each phase

Phase 1: Approval cycles

Interest in AI tools is high, but compliance reviews take time.

Phase 2: Time and output quality

One content designer called it the “too much work trap”: “I can’t really spare the time to learn a new tool, especially one that might not help.” Another respondent found that AI is “not yet at the quality bar for design that speeds us up.”

Similarly, 62% of designers in the AI in Design Report 2026 say inconsistent output is their biggest challenge.

Phase 3: Keeping context in sync

A media company describes the problem, keeping context in sync across tools:

“We have yet to find an ideal solution that allows us to both work in Figma and seamlessly update the context layer, which is crafted from multiple sources of truth on different SSO-authenticated platforms.”

42% of designers in the AI in Design Report also cite a lack of product or brand context as a challenge of using AI in design work.

Where is AI in product copy heading? Closer to code

The clearest shift in our data is about where product work happens. Nearly one-third of companies described designers prototyping in Claude Design, Claude Code, Cursor, or developer environments.

“We use Cursor to build designs from Figma or just from the prompt alone. We use it to write copy, look for inconsistencies, and build within our component library.”

— Content Designer, consumer investing app

“Designers who use AI agents have been using primarily Claude and Figma Make to create design mock-ups or prototypes in Figma, or to create working prototypes in a developer environment.”

— Content Designer, automotive mobility software company

Industry research shows this shift accelerating:

  • Designers are moving into development. In Figma’s report, the share of designers participating in development nearly doubled in a year, from 21% to 41%, and developers doing design work rose from 44% to 60%. Over half of respondents say they’ve stopped using linear handoffs entirely.
  • Coding tools are now design tools. 76% of designers have used an AI coding tool, and the use of AI for code generation rose by 31 points in a year.
  • Content design is getting more technical, too, closing the gap. In UXCC’s report, 66% of content designers say technical demands have increased, and 48% are expected to work with Git or GitHub.

Why this matters for product copy

AI has disrupted how teams build together. The handoffs, shared review spaces, and approvals that kept product copy consistent are disappearing. Copy is now created by more people, in more tools: 70% of designers use AI for UI copy and content writing, and respondents in our survey named more than 15 different AI tools.

Homegrown fixes don’t keep up. Each skill, gem, or custom agent holds its own copy of the rules, in its own tool, and keeping them in sync becomes another job. Nielsen Norman Group found that AI made individual tasks faster, but the time saved was often spent cleaning up other people’s AI-generated work.

That’s why we believe the next phase is shared infrastructure: one source of truth for product copy, connected to code, that every person and agent works from. And teams shouldn’t have to build it themselves. That’s what we’re building at Ditto. Read how we’re building for product workflows of the future.

The bottom line

AI in product copy is making real — but uneven — progress. Most teams are at chat tools or homegrown agents, and constraints like regulation determine how fast they move. The next challenge is giving every person and agent the same context as work moves closer to code.

We’ll share more analysis from this survey in the coming weeks. Subscribe to our newsletter for updates.

Methodology and notes

  • How we ran the survey: We emailed an open-ended survey to about 130 Ditto customers and collected responses from 32 people at 25 companies between September 22 and October 7, 2026.
  • What we asked: Whether respondents use AI agents in their work and in design, which agents and how, and, if not, whether it’s a personal or company decision.
  • Small, self-selected sample: Respondents are Ditto customers, and most champion Ditto within their companies. People with strong views on AI were likely more motivated to answer.
  • “Agent” meant different things: Some “yes” answers describe chat assistants rather than agentic workflows. We note the distinction where respondents made it. For definitions, see AI and agents 101: the basics for content designers.
  • Anonymized: We describe companies by their industry. One company asked that its written answers not be shared; we only used its responses in the counts.